Artem Butusov Blog
sormy.nm.ru
github.com/sormy
linkedin.com/in/sormy

LibreELEC as Time Machine backup network device

Author Artem Butusov · Published on June 1st, 2017

Contents

Why?

If you already have LibreELEC with big disk connected to it, maybe even RAID1 (mirror) which is really good for storing backups in safer way, then it could be a good idea to make LibreELEC works as cheap alternative to Apple Time Capsule.

Prerequisites

Install entware-ng (optware-like with opkg package manager) on your LibreELEC device firstly. Follow this article: https://www.artembutusov.com/libreelec-entware-ng-installation/

Installation

Install netatalk

Please note, this manual is for netatalk v3.x, there is a difference in configuration between v2.x and v3.x. See here for more details: http://netatalk.sourceforge.net/3.1/htmldocs/upgrade.html

Connect over SSH.

opkg install netatalk

Configure netatalk

Create folder for Time Machine backups:

mkdir -p "/storage/mirror/Time Machine"

Make sure that storage path is accessible for user who will access it, for anonymous access you will need to give “rwx” for all for target directory and all parents should also have “x” for all.

nano /opt/etc/afp.conf

Example with anonymous access:

[Global]
uam list = uams_guest.so

[Time Machine]
time machine = yes
path = /storage/mirror/Time Machine
vol size limit = 1048576
ea = none

[Mirror]
path = /storage/mirror
ea = none

Let anonymous (nobody) read afp.conf:

chmod 644 /opt/etc/afp.conf

Create systemd service units

Create service unit for afpd:

nano /storage/.config/system.d/netatalk.service

Please note, I set storage-mirror.mount in After to make sure that netatalk will start after my RAID1 will be mounted to /storage/mirror.

[Unit]
Requires=network-online.service
After=avahi-daemon.service storage-mirror.mount

[Service]
Type=oneshot
ExecStart=/opt/etc/init.d/S27afpd start
ExecStop=/opt/etc/init.d/S27afpd stop
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target

Enable netatalk on reboot and start service:

systemctl enable netatalk
systemctl start netatalk

Create service unit for cnid_metad:

nano /storage/.config/system.d/netatalk-cnid_metad.service

[Unit]
Requires=network-online.service
After=netatalk.service

[Service]
Type=oneshot
ExecStart=/opt/etc/init.d/S26cnid_metad start
ExecStop=/opt/etc/init.d/S26cnid_metad stop
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target

Enable cnid_metad on reboot and start service:

systemctl enable netatalk-cnid_metad
systemctl start netatalk-cnid_metad

Fix avahi configuration

Avahi daemon is used for auto discovery. Properly configured Samba and AFP shares will be available for mounting directly from OS X Finder.

There is only one problem. Both Samba and AFP shares could not work together on device with the same name. At least it doesn’t work well for me.

Here is a workaround: show AFP and Samba shares under different device name. In that case you will be able to mount both types of shares and auto discovery will still work for them.

There is one small problem actually. We need to edit avahi configuration, which is located in /etc on read only root.

We could either modify LibreELEC image OR we could use overlay fs to overlay etc. I found that second solution is easier.

Overlay /etc

Create folders for overlay fs:

mkdir -p /storage/modroot/etc /storage/modroot/.work

Create new systemd mount unit:

nano /storage/.config/system.d/etc.mount

with content:

[Unit]
DefaultDependencies=no
Conflicts=umount.target
Before=local-fs.target umount.target

[Mount]
What=overlay-etc
Where=/etc
Type=overlay
Options=lowerdir=/etc,upperdir=/storage/modroot/etc,workdir=/storage/modroot/.work

[Install]
WantedBy=local-fs.target

Enable and start /etc overlay mount:

systemctl enable etc.mount
systemctl start etc.mount

Now you should be able to create new files in /etc.

Create AFP avahi service

Create new file:

nano /etc/avahi/services/netatalk.service

with content:

<?xml version="1.0" standalone='no'?><!--*-nxml-*-->
<!DOCTYPE service-group SYSTEM "avahi-service.dtd">
<service-group>
    <name replace-wildcards="yes">AFP on %h</name>
    <service>
        <type>_afpovertcp._tcp</type>
        <port>548</port>
    </service>
    <service>
        <type>_device-info._tcp</type>
        <port>0</port>
        <txt-record>model=Xserve</txt-record>
    </service>
</service-group>

Create Samba avahi service

Create new file:

nano /etc/avahi/services/smb.service

with content:

<?xml version="1.0" standalone='no'?>
<!DOCTYPE service-group SYSTEM "avahi-service.dtd">
<service-group>
    <name replace-wildcards="yes">SMB on %h</name>
    <service>
        <type>_smb._tcp</type>
        <port>445</port>
    </service>
    <service>
        <type>_device-info._tcp</type>
        <port>0</port>
        <txt-record>model=Xserve</txt-record>
    </service>
</service-group>

Notes:

Restart avahi

Restart avahi:

systemctl restart avahi-daemon

Once you will restart avahi OS X Finder should show new devices and their shares.

Fix OS X configuration

By default OS X will not show our AFP shares as available for Time Machine backup but we could fix that.

Run in terminal on OS X:

defaults write com.apple.systempreferences TMShowUnsupportedNetworkVolumes 1

Usage

With this setup you should be able to see separate device in Finder for Samba and AFP shares. AFP device should have two mount points: Mirror and Time Machine. “Mirror” could be used for regular data and “Time Machine” should be available as backup target volume in OS X Time Machine backup tool.

You need to mount “Time Machine” volume first before it will become available in Time Machine as target backup disk.

In this setup “Time Machine” has no authentication and should be mounted with “Guest” account.

We could setup authorization but it will require modification of /etc/passwd. It is doable but will require more effort.

NOTE: Backing up to network device could be significantly slower due to slower IO over network and due to the fact that network backups are using sparse bundle drive images as container, which is slower than backing up directly to attached physical disk.

Extra documentation